Anatomical Balance of the Upper Airway and Obstructive Sleep Apnea

The investigation was approved by the institutional review board of Graduate School of Medicine, Chiba University, Chiba, Japan. Informed consent was obtained from each subject after the aim and potential risks were fully explained to each.

Two groups of adult male subjects (control group: 55 non-OSA subjects; OSA group: 50 OSA patients) were selected for this study after matching age, height, and maxillomandible dimensions. Exclusion criteria for both groups included presence of hypertrophied tonsils, acromegalics, and difficulty in identifying tongue silhouette on the cephalogram. The subjects were limited to men in this study because of possible craniofacial differences between sexes. OSA patients with an apnea–hypopnea index less than 20 episodes per hour were not included in the OSA group.

Obstructive sleep apnea patients were recruited from our clinic for sleep-disordered breathing. Because craniofacial abnormalities are common in OSA patients,11,12we considered that recruitment of potential non-OSA subjects from the general adult population would fail to match their craniofacial dimensions for those of OSA patients. Accordingly, 55 non-OSA male subjects were selected from three different populations. Because craniofacial abnormalities are known to be associated with difficult tracheal intubation,1817 male patients with suspicion or evidence of difficult tracheal intubation during anesthesia induction (Cormack and Lehane grade greater than 2 with laryngeal pressure) were included. Twenty-three male subjects who were referred to our clinic for preoperative airway assessment and were confirmed for absence of sleep-disordered breathing were also included. Furthermore, data from 15 non-OSA subjects previously reported were reanalyzed as control subjects with normal craniofacial characteristics.12 

Diagnosis of OSA was confirmed by standard full polysomnography, which was performed at one of the two local sleep laboratories (Komagamine Sleep Respiratory Center in Tokyo or Pulmonary Department of Sannou Hospital Medical Center in Chiba). Recordings include bilateral electroencephalograms, bilateral electrooculograms, submental electromyogram, leg electromyograms, electrocardiogram, airflow measurement with a thermistor at the mouth and nose, thoracoabdominal wall motions, oxygen saturation (Spo²), snoring over a microphone, and body position. Apnea was defined as absence of airflow for more than 10 s. Hypopnea was determined upon an apparent reduction of airflow for more than 10 s with reduction of Spo²by more than 4% from the baseline. Apneic events were classified as obstructive, mixed and central, and the apnea–hypopnea index was calculated as the total number of the obstructive or mixed apnea and hypopnea events per hour of sleep.

Nocturnal oximetry by a pulse oximeter (Pulsox-5 or Pulsox-3i; Minolta, Tokyo, Japan) and careful assessment of clinical symptoms were performed to exclude patients with potential sleep-disordered breathing from the control group. All potential control subjects were instructed to attach an oximetry finger probe before sleep and to remove the probe upon awakening. Digital readings of Spo²and pulse rate were stored every 5 s in the oximeter. The stored data were displayed on a computer screen to check quality of the recordings. The oxygen desaturation index, defined as the number of oxygen desaturation exceeding 4% from the baseline, the percentage of time spent at oxygen saturation less than 90%, the mean nadir of oxygen desaturations, and lowest Spo²were calculated by the computer. Nocturnal oximetry was repeated twice, and only subjects who presented normal oximetry—defined as an oxygen desaturation index <5 h⁻¹, the percentage of time spent at oxygen saturation less than 90% < 1%, and absence of daytime sleepiness—were included in the control group.18Thirteen habitual snorers without daytime sleepiness met the normal oximetry criteria and were included in the control group.

A lateral cephalometric radiograph was obtained for each patient in the upright position with natural head posture using a pair of earpieces (Chiba University Hospital). Before exposure, the subject was instructed to close the jaw in a natural occlusive position and to breathe quietly. The radiograph was taken at the end of expiration. The exposure parameters were arranged to clearly visualize bony landmarks.

Blind cephalogram analyses were made by an investigator (S.T.). Cephalometric parameters reflecting position and size of the maxilla and mandible were selected. As illustrated in figure 1, the following 10 cephalometric variables were measured and calculated for craniofacial characterization and estimation of tongue size.

Relative position of the mandible to the maxilla:  ANB angle = angle between the line from A (subspinale: the deepest point on the anterior surface of the maxillary alveolar bone) to N (nasion) and the line from B (supramentale: the deepest point on the anterior surface of the mandibular alveolar bone) to N.

Maxillomandible dimensions and lower face cage: 

1. Maxilla length (Cd-A) = distance between the medial condylar point of the mandible (Cd) and point A

2. Mandible length (Cd-Pog) = distance between Cd and pogonion (Pog: the most prominent point of the anterior surface of the mandibular symphysis in respect to the mandibular plane)

3. Developmental direction of the mandible: angle between Cd-A and Cd-Pog (angle A-Cd-Pog)

4. Axial component of the mandible length (Cd′-Pog) = mandible length projected to Cd-A line calculated as Cd-Pog · cos(A-Cd-Pog)

5. Longitudinal component of the mandible length (Cd-Cd′) = mandible length projected to vertical line to Cd-A calculated as Cd-Pog · sin(A-Cd-Pog)

6. Size of lower face cage–cross-sectional area (LF-CSA; dotted trapezoid in fig. 1): CSA of a trapezoid enclosed by Cd-A-Pog-Cd′ for assessment of bony enclosure size of the upper airway.

Hyoid position:  distance MP-H for evaluation of position of the hyoid bone = perpendicular distance from anterosuperior point of the hyoid bone (H) to mandibular plane (MP).

Tongue size:  cross-sectional area (T-CSA: shaded area in fig. 1) outlined by the dorsum of the tongue surface and lines that connect TT (tongue tip), RGN (retrognathion), H, and Eb (base of epiglottis).

Difference of the upper airway anatomical balance was assessed by comparison of the tongue size between craniofacial dimension–matched control and OSA groups. Because of wide range of craniofacial dimensions, the upper airway anatomical balance was also assessed in each of the following three subgroups with different craniofacial dimensions: small LF subgroup (LF-CSA ≤65 cm²), intermediate LF subgroup (65 cm²< LF-CSA < 71 cm²), and large LF subgroup (LF-CSA ≥71 m²). Furthermore, the tongue size was compared between body mass index (BMI) and craniofacial dimension–matched control and OSA subgroups.

Statistical analyses were performed by using a computer software (SigmaStat for Windows Version 3.11; Systat Software, Inc., Point Richmond, CA). Statistical differences between the control and OSA groups were assessed by Mann–Whitney rank sum test. Comparison between the subgroups was performed by Kruskal–Wallis one-way analysis of variance on ranks. All pairwise comparisons were performed by the Dunn method. A value of P < 0.05 was considered to be significant. All values are expressed as median (10th/90th percentiles).

Anthropometric characteristics and results of the sleep study are presented in table 1. Age and height did not differ between the groups. OSA patients were more obese and had larger neck circumference than control subjects. Polysomnogram revealed the presence of severe OSA accompanied with severe desaturations in OSA patients. Absence of sleep-disordered breathing was confirmed by few episodes of nocturnal desaturations in control subjects.

Results of cephalometric analyses are presented in table 2. Relative position of the mandible to the maxilla and maxillomandibular dimensions were heterogeneous, ranging widely in both the control and OSA groups, and did not differ between the groups. The size of the lower face cage did not differ between the groups, verifying matching craniofacial dimensions and bony enclosure size surrounding the upper airway between the groups.

Compared with the craniofacial dimension–matched non-OSA subjects, a significantly larger tongue was evident in OSA patients, indicating anatomical imbalance surrounding the upper airway (table 2). Results indicate a more caudal location of the larger tongue in OSA patients, expanding from the lower face cage as evident from a significantly longer MP-H distance. These results are clearly demonstrated in figure 2, which shows non-OSA and OSA subjects with equivalent lower face size but different tongue CSA.

Results of subgroup cephalometric analyses are presented in table 3. While the maxillomandible dimensions were significantly larger in the larger LF subgroups by the definitions, the dimensions did not differ between non-OSA and OSA patients within each of the subgroups, indicating successful matching of craniofacial dimensions in the subgroups. While the tongue was larger in subjects with larger lower face cage, the tongue was significantly larger in OSA patients than non-OSA subjects by 3–5 cm², indicating excessive upper airway soft tissue for a given craniofacial size in OSA patients (fig. 3). Noticeably, the tongue of non-OSA subjects with larger lower face cage did not differ from that of OSA patients with smaller lower face cage, suggesting little significance of absolute tongue size and significance of imbalance of tongue size and lower face cage for development of OSA. Interestingly, MP-H distance did not significantly differ between non-OSA and OSA in the large LF subgroup.

By selecting BMI-matched non-OSA subjects and OSA patients from the LF-CSA matched subgroups, we succeeded in matching BMI and craniofacial dimension between 13 non-OSA and 13 OSA patients (table 1 and 2). Neck circumference was longer in the OSA subgroup than in the non-OSA subgroup. A larger tongue and a more caudal location of the hyoid were evident in OSA patients.

This is the first study that compares tongue size between craniofacial dimension–matched non-OSA subjects and OSA patients, including wide range of craniofacial characteristics from normal to abnormal. While the tongue was significantly larger in subjects with larger maxillomandible dimensions, OSA patients had a significantly larger tongue for a given maxillomandible size, suggesting that the anatomical imbalance surrounding the upper airway contributes to development of OSA.

Our structural analysis was two-dimensional and did not include the whole upper airway structure possibly involved in development of OSA, indicating a major methodologic drawback for investigation of OSA pathophysiology, unlike sophisticated three-dimensional analyses.16,17Nevertheless, it was surprising to find upper airway anatomical imbalance despite the methodologic limitation. Although our approach may have an advantage for clinical applicability and usefulness over three-dimensional analyses, future studies including volumetric three-dimensional analyses are necessary to provide conclusive evidence for upper airway anatomical imbalance as an OSA pathogenesis.

Another major limitation is the potential bias in selecting non-OSA control subjects. Polysomnography would have been preferable to exclude OSA from potential control subjects, because subjects with craniofacial abnormalities known to be associated with OSA were actively included.6–11However, control subjects were recruited from three different populations and do not represent a true non-OSA population. However, recognizing this potential bias, we invited subjects with difficult tracheal intubation to increase the number of control subjects with craniofacial abnormalities.19In reality, it is difficult to obtain numerous control subjects with craniofacial abnormalities from the general adult population to match craniofacial dimensions between the groups and test our hypothesis.

The study population was limited to Japanese adult men. Because of the differences of BMI and craniofacial characteristics between sex and race, our results may not be completely applicable to females and other races. White OSA patients are more obese and have larger maxillomandible dimensions than Asian OSA patients,20,21agreeing with the upper airway anatomical balance concept across races. However, obesity may contribute to the development of OSA by reduction of lung volume in addition to excessive upper airway soft tissue.22Differences of respiratory chemosensitivity23and reflex control of the upper airway muscles24among the subjects were not assessed and controlled in this study, which may have produced a significant overlap of tongue size between non-OSA and OSA for a given craniofacial dimension clearly evident in figure 3.

As indicated by significant association between tongue size and lower face cage, absolute tongue size alone may contribute to development of OSA less than the relative tongue size representing balance between the amount of soft tissue and bony enclosure size surrounding the upper airway. Shelton et al.  25reported that both mandible enclosure size and body weight are important determinants of the number of OSA episodes. We extended their concept and confirmed that anatomical balance between the amount of soft tissue inside the maxillomandibular enclosure and the bony enclosure size determines the pharyngeal airway size.12,13,26Anatomical imbalance is not usually evident during wakefulness and pharyngeal airway patency is maintained mainly by neural compensatory mechanisms in OSA patients.24The anatomical balance model presented in figure 4may be helpful for conceptualizing and understanding the interaction between neural and anatomical mechanisms, while it ignores numerous factors that may influence pharyngeal airway patency.12,13,27 

Although heterogeneous craniofacial features among humans may be determined by both genetic and environmental influences, development of craniofacial structures is generally accomplished by 20 yr of age, when prevalence of OSA is low.28Prevalence of both obesity and OSA increases during middle age.29Because one of the two anatomical balances, i.e. , bony enclosure size, in figure 4is relatively constant during adulthood, our results suggest that OSA could develop in subjects with excessive upper airway soft tissue caused possibly by being overweight.

Our results completely agree with those recently reported by Schwab et al. ,17who performed magnetic resonance imaging three-dimensional volumetric measurements of soft tissue surrounding the upper airway in both control and OSA subjects. After statistical adjustments for sex, race, age, craniofacial size, and parapharyngeal fat, they estimated significant differences of tongue and total upper airway soft tissue between control and OSA subjects by 17 cm³and 21 cm³, respectively. It should be noted that the difference is close to the 5-cm²difference of tongue CSA between control and OSA subjects found in this study, assuming both maxillary and mandibular arch width to be approximately 3–5 cm.30In addition to confirming their results, we first demonstrated interaction between the craniofacial structures and soft tissue surrounding the pharynx for determining OSA development in craniofacial dimension–matched non-OSA and OSA groups.

Although the maxillomandible bony enclosure and the cranial base limit soft-tissue expansion, the submandibular area does not have the bony structure of this region, thereby permitting caudal expansion of the excessive soft tissue. The hyoid bone was demonstrated to be located more caudally in OSA patients in accord with previous studies.11,12Because the human hyoid bone has no connection to other bones and is mobile, caudal expansion of soft tissue resulting from excessive soft tissue within the maxillomandible enclosure may be responsible for shifting the hyoid bone caudally. It is our belief that caudal displacement of the hyoid bone reflects upper airway anatomical imbalance in OSA patients.

While a variety of physical examinations are proposed as potential predictors of OSA, modified Mallampati score, thyromental angle, and cricomental space are reported to be better differentiating variables of OSA and non-OSA.31,32The results of this study provide a pathophysiologic explanation for the usefulness because these assess relative excessiveness of soft tissue surrounding the upper airway and agree with the anatomical balance model. Furthermore, our cephalometric analysis could serve as an alternative preoperative predictor by measuring both maxillomandibular dimension and tongue CSA in addition to position of the hyoid bone.

Within one subject, the pharyngeal bony enclosure size varies with head and mandible positioning changes, possibly improving the anatomical imbalance observed in OSA patients. Head extension, the sniffing position, and mandible advancement increase the distance between the mentum and the cervical column increasing the bony enclosure size. We previously demonstrated significant improvement of pharyngeal closing pressures in response to these position changes in anesthetized and paralyzed OSA patients.13,26,33Furthermore, the sitting position may improve the anatomical balance by displacing the excessive soft tissue outside the bony enclosure through the submandibular space. The upper airway anatomical balance concept may provide the base for establishing strategies for perioperative airway management of OSA patients.

This study demonstrated that OSA patients have excessive soft tissue for a given craniofacial dimension. An upper airway anatomical balance model was proposed for better understanding the pathophysiology of pharyngeal obstruction and OSA.

The authors thank Sara Shimizu, M.D. (Head of the Department of Plastic Surgery, JFE Kawatetsu Chiba Hospital, Chiba, Japan), who greatly helped to improve the manuscript.